Expected goals

[desmondwolff]

Does FM/SI have any plans to introduce the Expected Goals metric into the next game?

Expected Goals: http://11tegen11.net/2014/02/10/what-is-expg/

http://cartilagefreecaptain.sbnation.com/2015/10/19/9295905/premier-league-projections-and-new-expected-goals

[tyler16]

I'd love it if they did.

[ac13]

I think it would show up the match engine too much. The expected goals would be + a lot for every team because the chances created in game are from much closer range than irl. Most opposition teams barely have a shot from outside the area and very few are even from the 12-18 yard mark of the box. Far too many are from close range.

[Svenc]

I think it would show up the match engine too much. The expected goals would be + a lot for every team because the chances created in game are from much closer range than irl. Most opposition teams barely have a shot from outside the area and very few are even from the 12-18 yard mark of the box. Far too many are from close range.

Quite funny considering that there's always the complaints of too many shots from range. Could actually well be, it'S a computer sim TARGETING football, but obviously never reaching it. Out of interest, how did you come to that conclusion? The chalkboard if you click it on whoscored for instance shows the shot spreads for each match: https://www.whoscored.com/Matches/959867/Live/England-Premier-League-2015-2016-Manchester-City-Manchester-United And various sides collect data based on this and similar. However in-game there's none of such metrics available, you'd have to do them all manuably. I suspect too by the way that having more detailed ones would reveal additionally areas where FM is actually hugely off football, i.e. such as the number of tackles. Not all necessarily due to stuff inherent to the ME, but also because of how teams apply tactics. For human players it's comparably attackattackattack on average, with many never developing a sense increasing and decreasing risk during a match, and conversely the AI at its most rigid it's totally shutting up shop very readily with like no player advancing outside of counter attacks and set pieces visibly. Both firstly rubs off on match play first and foremost, which then rubs off on statistics. It is the visible play, second by second making the numbers, not the other way around ever. Still curious how you did this.

Not sure about ExpG in the game, there's more advanced analysis that deals in game states, as obviously in football goals depend on each other some, therefore expectations shift with scorelines and with minutes having passed (i.e. tactical changes after teams taking leads or going behind on both sides typically, outright late game desperataly overloads vs. shutting up shop for good but also fatigue kicking in). But anything that would be a better metric of quality rather than the dreaded "clear cuts" would be an improvement as for the stats department, which some sadly focus on completely in isolation, when numbers will never show the fully picture. It's clear it's not working, and it's apparent why it can't, even taking into account that classifying chances is a subjective thing in real life done by actually human eyes in real life football analysis. You can feed a stat algorithm with as many factors as you like, distance to goal, distance of next defender near, angle of shot, wrong or right foot, position of keeper, all more obvious factors in FM too etc. It will always pick up on stuff that isn't all that and vice versa. Additionally SI doesn't even reveal what the expected average conversion should be like in their categories listed. Therefore the added frustration of some thinking that a clear cut should result in a goal in 90% of times, whereas in real football even outright penalties aren't converted than at a higher rate of 80% on average, and Opta clear cuts (typically one on ones based on their definition) are converted at rates in between 1 in 3 to 1 in 5, and the difference long-term between the average and the world class isn't that comically big.

[s1111]

Expected goals would certainly be preferable to the vague clear cut/half chance stat currently in the game. The problem would be measuring it accurately in a decimal metric when the game already struggles defining what and how good a chance is in its current vague system.

And all the whiners posting on here saying 'I had 1.8 xG and the AI 0.7 but I lost 2-0 why is the game cheating' would be unbearable.

[desmondwolff]

All good points!

My thinking was twofold:

1 -- personally, i'd like it

2 -- the real game is already using it (even old man Wenger talks about ExG!)

[alian62]

Well its not accurate so why would they. Going on the table you linked to the Top Four finishers did not include Tottenham or Leicester

[desmondwolff]

It's an emerging metric that is outperforming others. That aside, the bigger argument for including it is that the real world game really is including it, irrespective of its merits. To not include it on grounds we don't like it would be like saying we won't include referee personalities since we don't like the fact that real world refs have personalities!

[forameuss]

It's an emerging metric that is outperforming others. That aside, the bigger argument for including it is that the real world game really is including it, irrespective of its merits. To not include it on grounds we don't like it would be like saying we won't include referee personalities since we don't like the fact that real world refs have personalities!

Not quite as simple as that though. It's a statistical metric, one of hundreds that I'm sure are used in some way that aren't available in FM. Should they be? Some of them probably. But then I could draw up a metric that shows where teams should be finishing based on the colour of their strips. Doesn't mean it should be in the game just because it exists.

Plus it doesn't really seem that useful to me. Looking at that table they've created using it, it's not showing a massively different table, but obviously there are some big exceptions. So what does it really tell you? I get that some people love their statistics, but when teams can soak up pressure for 90 minutes, and then benefit from a freak own goal to take all three points, where does this statistic tell you anything useful?

[desmondwolff]

Not quite as simple as that though. It's a statistical metric, one of hundreds that I'm sure are used in some way that aren't available in FM. Should they be? Some of them probably. But then I could draw up a metric that shows where teams should be finishing based on the colour of their strips. Doesn't mean it should be in the game just because it exists.

Plus it doesn't really seem that useful to me. Looking at that table they've created using it, it's not showing a massively different table, but obviously there are some big exceptions. So what does it really tell you? I get that some people love their statistics, but when teams can soak up pressure for 90 minutes, and then benefit from a freak own goal to take all three points, where does this statistic tell you anything useful?

You're right to say anyone can come up with any nonsense metrics! The difference is that people who do this for a living are showing that this metric out-performs all others (including possession, shots on target, etc).

Your example: ExG shows you that if you soak up pressure for 90 minutes and then concede on the break, yet also have a very high expected shots count, you shouldn't change anything because over multiple games you'll win. This guards against the chance that a manager might over-correct just because they've conceded on the counter in one game. This is one of the reasons some clubs are getting into it.